Holocene atmospheric circulation in the central North Pacific: a new terrestrial
1
diatom and δ18O dataset from the Aleutian Islands
2 3
Hannah L Bailey a,b,*, Darrell S Kaufman c, Hilary J Sloane d, Alun L Hubbard e, f, Andrew 4
CG Henderson g, Melanie J Leng d, h, Hanno Meyer b, and Jeffrey M Welker a 5
6
a Department of Biological Sciences, University of Alaska Anchorage, Anchorage, AK 99508, USA 7
bAlfred Wegener Institute for Polar and Marine Research, Potsdam 14473, Germany 8
c School of Earth Sciences & Environmental Sustainability, Northern Arizona University, Flagstaff, AZ 86011, 9
10 USA
d NERC Isotope Geosciences Facility, British Geological Survey, Nottingham NG12 5GG, UK 11
eCentre for Arctic Gas Hydrate, Environment and Climate, Department of Geology, UiT The Arctic University 12
of Norway, 9037 Tromsø, Norway 13
fInstitute of Geography & Earth Sciences, Aberystwyth University, Aberystwyth SY23 3DB, UK 14
g School of Geography, Politics and Sociology, Newcastle University, Newcastle-upon-Tyne, NE7 1RU, UK 15
hCentre for Environmental Geochemistry, School of Biosciences, University of Nottingham, Loughborough, 16
LE12 5RD, UK 17
* Corresponding author. Email address: [email protected] 18
19
Key words: Holocene; Paleoclimate; North Pacific; Limnology; Stable Isotopes; Diatoms 20
21
Highlights:
22
▪ New Holocene oxygen isotope record from the Aleutian Islands 23
▪ Diatom δ18O reflects shifts in synoptic-scale atmospheric circulation 24
▪ Warmer/wetter early-mid Holocene, cooler/drier after 4.5 ka 25
▪ Enhanced winter circulation corresponds to Holocene glacier advances 26
▪ Current environmental changes unprecedented within past 9.6 ka 27
Abstract 28
The North Pacific is a zone of cyclogenesis that modulates synoptic-scale atmospheric 29
circulation, yet there is a paucity of instrumental and paleoclimate data to fully constrain its 30
long-term state and variability. We present the first Holocene oxygen isotope record 31
(δ18Odiatom) from the Aleutian Islands, using siliceous diatoms preserved in Heart Lake on 32
Adak Island (51.85° N, 176.69° W). This study builds on previous work demonstrating that 33
Heart Lake sedimentary δ18Odiatom values record the δ18O signal of precipitation, and correlate 34
significantly with atmospheric circulation indices over the past century. We apply this 35
empirical relationship to interpret a new 9.6 ka δ18Odiatom record from the same lake, 36
supported by diatom assemblage analysis. Our results demonstrate distinct shifts in the 37
prevailing trajectory of storm systems that drove spatially heterogeneous patterns of moisture 38
delivery and climate across the region. During the early-mid Holocene, a warmer/wetter 39
climate prevailed due to a predominantly westerly Aleutian Low that enhanced advection of 40
warm 18O-enriched Pacific moisture to Adak, and culminated in a δ18Odiatom maxima (33.3 ‰) 41
at 7.6 ka during the Holocene Thermal Maximum. After 4.5 ka, relatively lower δ18Odiatom
42
indicates cooler/drier conditions associated with enhanced northerly circulation that persisted 43
into the 21st century. Our analysis is consistent with surface climate conditions inferred from 44
a suite of terrestrial and marine climate-proxy records. This new Holocene dataset bridges the 45
gap in an expanding regional network of paleoisotope studies, and provides a fresh 46
assessment of the complex spatial patterns of Holocene climate across Beringia and the 47
atmospheric forces driving them.
48 49 50 51 52 53
1. Introduction 54
Numerous paleoenvironmental studies now contribute to a global synthesis and 55
understanding of Holocene climate change over the past 11.7 ka [Mayewski et al. 2004;
56
Marcott et al. 2013; Rehfeld et al. 2018]. By comparing common trends between individual 57
proxy records, these studies provide a means to infer the timing, scale, and spatial extent of 58
major Holocene climatic features. These include stepwise climate transitions, intervals 59
exceeding twentieth century warmth, and the low-frequency behaviour and modes of natural 60
climate variability. At broad (i.e. global) spatial and temporal scales these trends are 61
relatively coherent and unambiguous, yet at finer spatial scales, climate variability is more 62
pronounced due to local and regional factors. Such variability is highlighted in two recent 63
paleoclimate syntheses focused on west and eastern Beringia – the region extending from 64
northeast Siberia to northwest Canada (Fig. 1a) [Brooks et al. 2015; Kaufman et al. 2016].
65
While general circulation models (GCM) typically emphasise insolation as the key driver of 66
millennial-scale Holocene climate change [Renssen et al. 2009], these compilations indicate a 67
more complex and spatially heterogeneous climate evolution than implied by linear insolation 68
forcing alone. For example, major climatic features previously considered ubiquitous, such as 69
a prominent Holocene thermal maximum (HTM) [Kaufman et al. 2004], are now recognised 70
to be spatially asynchronous across this vast region [Kaufman et al. 2016]. Moreover, 71
existing terrestrial water isotope records are also shown to be ambiguous and contradictory 72
during the Holocene [Kaufman et al. 2016] and the most recent suite of model-data 73
comparisons reveal significant mismatches between simulated and reconstructed Holocene 74
temperatures in Alaska [Zhang et al. 2017].
75
At a synoptic scale, Beringia is located within the main centre of influence of the 76
Aleutian Low, one of the most dominant ocean-atmospheric systems in the Northern 77
Hemisphere with global climatic significance [Rodionov et al. 2007]. However, virtually all 78
available terrestrial paleoclimate data are restricted to mainland Alaska and eastern Russia 79
[Sundqvist et al. 2014; Brooks et al. 2015; Kaufman et al. 2016], and compared to lower 80
latitude regions, paleoisotope reconstructions are sparse [Kaufman et al. 2016]. This partly 81
reflects a lack of base-line water isotope measurements for constraining the regional water 82
isotope cycle [e.g. Welker, 2000; Anderson et al. 2016], as well as a paucity of lake core 83
studies with continuous sequences of carbonate-rich sediments – or suitable alternatives − for 84
isotopic analysis. Hence, to elucidate past and future climate in this region, there is an 85
outstanding requirement for greater spatial coverage of highly resolved and accurately dated 86
paleoclimate datasets, as well as an empirical-based understanding of the atmospheric and 87
environmental controls driving them.
88
To address this, we present the first Holocene oxygen isotope record from the 89
Aleutian Islands in south west Alaska. Our isotope measurements derive from siliceous 90
diatoms (δ18Odiatom) preserved in the sediments of Heart Lake, on Adak Island (Fig. 1b), and 91
are supported by diatom assemblage analysis of the same sedimentary sequence. We build on 92
earlier work by Bailey et al. [2015] who demonstrate that Heart Lake δ18Odiatom values 93
correlate significantly with North Pacific climate indices over the past hundred years (r = 94
0.43; p < 0.02, n = 28). Here, we apply this empirically-derived understanding to interpret 95
new δ18Odiatom data from a longer Heart Lake sediment core which extends back to 9.6 ka.
96
The primary aims are to: (1) investigate the forcing and response of this remote region to a 97
warming climate system as it transitioned from the last glacial period; (2) develop a Holocene 98
reconstruction of North Pacific atmospheric circulation; and (3) bridge the gap in the regional 99
network of proxy records to synthesise and assess spatio-temporal patterns of natural climate 100
variability across Beringia.
101 102
2. Regional Setting 103
Heart Lake is a small (~0.25 km2), freshwater through-flow system on Adak Island in the 104
central North Pacific (51.85 ° N, 176.69 ° W) (Fig.1c). The island is volcanic and forms part 105
of the 1900-km-long Aleutian archipelago extending from mainland Alaska to the Russian- 106
Kamchatka Peninsula. The lake watershed area is ~8 km2 and is situated in low-relief hills 107
surrounded by mountainous terrain (Fig. 1c). There is a single lake basin with a maximum 108
depth of 8 m. One stream inflows from two larger lakes and a small outflow channel drains to 109
the Bering Sea ~2 km to the west. Lake volume is ~8 ×105 m3 and water retention is an 110
estimated two weeks, based on the available stream gauge inflow data [TDX, 2013].
111
Inspection of available satellite imagery reveals that Heart Lake freezes over in winter and 112
this ice surface remains into spring [USGS, 2017].
113
114
Figure 1. Location of (a) Adak Island in the central Aleutian Islands; (b) Heart Lake 115
and Andrew Lake; (c) oblique north west view of Heart Lake with the inflow channel 116
visible in the foreground [credit: Yarrow Axford]; and (d) monthly mean precipitation 117
(blue bars) and surface air temperature at Adak airport (1949−2016), whereby solid 118
lines depict mean (black), minimum (blue) and maximum (red) temperatures [NOAA, 119
2017]. Numbered circles in 1a indicate key sites referred to in text: (1) LV29-114-3 [Max 120
et al. 2012], (2) Pechora Lake [Hammarlund et al. 2015], (3) SO201-12-77KL [Max et al.
121
2012], (4) Horse Trail Fen [Jones et al. 2014], (5) Mica Lake [Schiff et al. 2009], (6) 122
Mount Logan [Fisher et al. 2008], and (7) Jellybean Lake [Anderson et al. 2005]
123
Adak Island has a mild maritime climate compared to mainland Alaska and is 124
strongly affected by persistent fog and light rain in the summer, and frequent storms and 125
strong winds during winter [Rodionov et al. 2007]. Mean annual air temperature is +4.3 °C, 126
and mean winter (December−February) and summer (June−August) values are +1.0 °C and 127
+9.0 °C, respectively (1949−2016) [NOAA, 2017]. Mean December and July precipitation is 128
163 mm and 71 mm, respectively (Fig. 1d) [NOAA, 2017]. Of the total 1.3 m annual 129
precipitation, ~75 % (1.0 m) falls from September to February.
130
The regional climate reflects the configuration of large scale atmospheric−ocean 131
systems, namely the Aleutian Low: a synoptic-scale feature of mean low sea level pressure 132
(SLP) and the leading driver of North Pacific climate[Mock et al. 1998]. When the Aleutian 133
Low is ‘weak’, storms tend to track north over the central Aleutian Islands (Fig. 2a); when 134
the pressure system is ‘strong’, storms track south of the Aleutians and into the Gulf of 135
Alaska (Fig. 2b) [Mock et al. 1998; Rodionov et al. 2007]. These circulation patterns vary on 136
interannual to decadal timescales and induce characteristic climate responses that are well 137
expressed in coupled modes of the North Pacific Index (NPI) and the Pacific Decadal 138
Oscillation (PDO)[Trenberth and Hurrell, 1994; Mantua et al. 1997]. Typically, a strong 139
Aleutian Low (−NPI/+PDO) will induce positive sea surface temperatures (SST), surface air 140
temperatures (SAT), and precipitation anomalies in the Gulf of Alaska and negative 141
anomalies in the central North Pacific, with contrary conditions during a weak Aleutian Low 142
(+NPI/−PDO) (see Supplementary Fig.1).
143
144
Figure 2. Mean winter (December−February) sea level pressure associated with the six 145
most positive (a) and negative (b) North Pacific Index (NPI) values between 1950 and 146
2017 [Trenberth and Hurrell, 2004]. A negative (positive) NPI is a strong (weak) 147
Aleutian Low. Arrows highlight the direction of the primary storm tracks delivering 148
precipitation to our site on Adak Island (yellow star) [Bailey et al. 2015]. SLP data 149
obtained from NCEP/NCAR V1 reanalysis[Kalnay et al. 1996]. Numbered yellow circles 150
in (a) indicate locations of the (1) LV29-114-3 [Max et al. 2012], (2) Pechora Lake 151
[Hammarlund et al. 2015], (3) SO201-12-77KL [Max et al.2012], (4) Horse Trail Fen 152
[Jones et al. 2014], (5) Mica Lake [Schiff et al. 2009], (6) Mount Logan [Fisher et al.
153
2008], and (7) Jellybean Lake [Anderson et al. 2005] climate records discussed in text.
154 155
3. Materials and Methods 156
3.1. Sediment and water recovery 157
Sediment cores and bottom lake water samples were recovered from Heart Lake during the 158
summers of 2009 and 2010. A Garmin GPS sonar was used to survey its bathymetry and 159
reveals that Heart Lake comprises of a single basin with a maximum depth of 8 m, 160
surrounded by a shallow platform < 2 m deep (see Supplementary Figure 2). Coring sites 161
were selected adjacent? (in the vicinity?) of the deepest part of the basin at a depth of 7.6 m.
162
Seven sediment cores were extracted using percussion and hand-held gravity coring devices 163
operated from a floating platform. Bottom lake water samples were collected in situ at the 164
sediment-water interface during gravity coring. Following core extraction water was 165
immediately siphoned and sealed in 50 ml vials, ensuring no head space. (why did water 166
sampling have to proceed immediately after core extraction? Were they not sampled using a 167
niskin sampler or similar?) Sediment cores were then split lengthways, packaged, and 168
shipped with water samples to Northern Arizona University where they were stored at 4°C 169
until they were sub-sampled and analyzed. Our study focuses on the longest percussion core 170
(10-AS-1D; 5.9 m) and two accompanying surface gravity cores (09-AS-1A, 0.81 m; and 09- 171
AS-1B, 0.44 m). For a detailed description of the sediment core’s lithostratigraphy, see 172
Krawiec et al. [2013].
173
3.2. Chronology 174
The composite age model for 10-AS-1D and 09-AS-1A is presented in a separate paper 175
devoted to the tephrostratigraphy and radiometric dating of the Heart Lake sedimentary 176
sequence [Krawiec et al. 2013]. In summary, a Monte Carlo approach was employed to 177
model the age-depth relation of 16 macrofossil AMS radiocarbon (14C) dates, together with a 178
peak in recent 239+240Pu activity and the age of the sediment-water interface (2009 AD) 179
[Krawiec et al. 2013]. Tephrostratigraphy was used to independently cross check the 180
accuracy of the chronology, whereby the ages of down core tephra horizons from Heart Lake 181
were compared with tephra ages from nearby Andrew Lake and previously published outcrop 182
studies [Krawiec et al. 2013]. The chronology for surface core 09-AS-1B derives from 183
radiometric dating of 210Pb, 226Ra, 137Cs and 241Am by direct gamma assay on 14 dried 184
sediment samples from the upper core section [Bailey et al. 2015]. The cores were cross- 185
correlated using a prominent tephra horizon found in all three sedimentary sequences 186
[Krawiec et al. 2013; Bailey et al. 2015]. All ages herein are expressed as thousands of 187
calendar years (ka) prior to 1950 AD, where 1 ka = 1000 cal yr BP.
188 189
3.3. Stable isotope analyses 190
A total of 147 sediment samples were processed for δ18Odiatom analysis. These samples range 191
in age from 9.6 ka (587 cm depth) to 2009 AD, and are sub-/decadally resolved for the most 192
recent 1500 years and at centennial resolution thereafter. From the 5.9 m long core 10-AS- 193
1D, 1 cm3 of sediment (i.e. a 1-cm-thick sample – this doesn’t quite make sense – it must 194
have been a very small diameter corer for 1 cm thickness to yield 1 cm^3 of sediment?) was 195
extracted at 7 cm intervals from the base (587 cm) to the top of the core. This was the optimal 196
sampling resolution to avoid tephra layers which could potentially cause contamination issues 197
[Lamb et al. 2007] – sounds slightly disingenuous as can’t imagine that >80% of the core was 198
tephra – why not leave this sentence out? The surface cores 09-AS-1A and 09-AS-1B were 199
both sampled in contiguous 0.5 cm increments. This detail was used to capture sub-decadal 200
changes in δ18Odiatom over the past century for direct comparison with instrumental records 201
[see Bailey et al. 2015]
202 203
Sediment samples were prepared using a hybrid process of chemical digestion, 204
sieving, and heavy liquid separation adapted from Morley et al. [2004]. To remove organic 205
and carbonate material, samples were treated with 30% H2O2 at 90°C until reactions ceased, 206
before using 5 % HCl at ambient temperature. Samples were then centrifuged in sodium 207
polytungstate (3Na2WO49WO3.H2O) (SPT) heavy liquid at 2500 rpm for 20 minutes, 208
resulting in the separation and suspension of diatoms from the heavier detritus. This 209
procedure was repeated three times for each sample using specific gravities of 2.50, 2.30 and 210
2.25 g ml-1. After the final SPT separation, samples were washed five times in ultrapure water 211
(UPW) at 1500 rpm for 5 minutes and vacuum filtered through a 3 µm cellulose nitrate 212
membrane to remove potential clay minerals and/or broken diatom fragments. The < 3 µm 213
fraction was discarded as it was too small (< 1 mg) to be analyzed and, upon further 214
inspection (need to put an ‘a’ in here if keeping this in) contained only small broken diatom 215
fragments and detritus. The remaining samples were treated with a final stage of 30 % H2O2
216
at 60 °C for one week to ensure no traces of organic matter remained.
217
Purified diatom samples were analyzed for δ18Odiatom using the stepwise fluorination 218
method [Leng and Sloane, 2008] at the NERC Isotope Geosciences Laboratory in Keyworth, 219
UK. The outer hydrous layer of the diatom was removed in a pre-fluorination stage using a 220
BrF5 reagent at low temperature [Leclerc and Labeyrie, 1987]. This was followed by a full 221
reaction at high temperature to liberate oxygen that was converted to CO2 [Clayton and 222
Mayeda, 1963] and measured for δ18Odiatom using a MAT 253 dual-inlet mass spectrometer.
223
Replicate analyses indicate an analytical reproducibility of ±0.19 ‰ (1σ) for the samples, and 224
±0.30 ‰ (1σ) for the diatom standard BFCmod. All δ18Ovalues were converted to the Vienna 225
Standard Mean Ocean Water (VSMOW) scale using the BFCmod standard for calibration.
226
Two Heart Lake water samples were measured for their oxygen and hydrogen (δD) 227
isotope composition using a Thermo-Finnigan Deltaplus XL gas mass spectrometer at the 228
Colorado Plateau Stable Isotope Laboratory, Northern Arizona University, USA. Analytical 229
precision on internal working standards was ±0.1 % for δ18O and ±1 % for δD. All values are 230
reported here in per mil (‰) relative to VSMOW.
231 232
3.3.1. Contamination assessment 233
All purified diatom samples (n = 147) were visually inspected for contamination using an 234
OLYMPUS BX40 light microscope. Thirty samples were selected down-core and further 235
inspected using a Hitachi S-4700 field emission scanning electron microscope (SEM). In 236
addition, fourier transform infrared spectroscopy (FTIR) was applied to assess the chemical 237
composition and sample purity of 16 diatom samples from core 10-AS-1D [Swann and 238
Patwardham, 2011]. These samples, together with the BFCmod diatom standard, were 239
analyzed using FTIR at the British Geological Survey in Keyworth, UK [Bailey et al. 2014].
240
FTIR analyses of all purified diatom isotope samples measured indicate peaks corresponding 241
to the BFCmod standard, known to represent clean, fossilised diatomite (Supplementary Fig.
242
3). Spectral deviation from the standard would indicate additional compounds and 243
contamination by non-diatom components [Swann and Patwardhan, 2011]; peaks centred at 244
~450 cm-1, ~800 cm-1 and ~1100 cm-1 confirm pure silica and the integrity of our diatom 245
isotope samples [Bailey et al. 2014].
246 247
3.4. Diatom assemblage analysis 248
Fifty-seven sub-samples of the purified diatom material used for δ18Odiatom analysis were 249
retained for diatom species analysis. These include 33 samples selected at c. 13 cm intervals 250
from AS-10-1D, and 24 samples at a contiguous 0.5 cm resolution from AS-09-1B. Diatom 251
slides were prepared on a hot plate using Naphrax® mounting medium. A minimum of 300 252
diatom frustules per sample were counted along transects at x1000 magnification, under an 253
OLYMPUS BX40 light microscope. Taxonomic identification was based on classifications in 254
Camburn and Charles [2000] and Krammer and Lange-Bertalot [1986−1991].
255
Following diatom identification, species counts were converted to percentage 256
abundance and evaluated using the software package Tilia (v.2.0.41) [Grimm, 2015]. For 257
diatom zone demarcation, a constrained incremental sum-of-squares cluster analysis 258
(CONISS) [Grimm, 1987] was applied to all dominant taxa with a relative abundance >5 % 259
in at least one sample. To quantitatively assess down core trends in diatom assemblages, a 260
principal components analysis (PCA) [ter Braak and Prentice, 1988] was applied to a 261
correlation matrix based on the dominant (>5 %) diatom species in all 57 samples. The 262
analysis was performed on untransformed percentage data using the program C2 (v.1.7.6) 263
[Juggins, 2014].
264 265
4. Results 266
4.1. Diatom flora 267
Diatom frustules are well preserved in all samples and show no sign of valve dissolution. The 268
flora is diverse and a total of 155 different freshwater diatom species were identified. Of 269
these, 11 species account for > 90 % of all diatoms present in all samples. These include 270
species belonging to the genera Aulacoseira, Cyclotella, Rossithidium, and small fragilarioid 271
taxa (consisting of the genera Fragilaria, Pseudostaurosira, Staurosira, Stauroforma, and 272
Staurosirella). Species with an abundance ≥ 5 % in at least one stratigraphic level are 273
presented (Fig. 3), and the record is divided into four zones based on the CONISS 274
dendrogram: Zone 1 (9.6−8.6 ka; 587−452 cm), Zone 2 (8.6−4.4 ka; 452−352 cm), Zone 3 275
(4.4 ka−1860 AD; 352−13.25 cm), and Zone 4 (1860−2009 AD; 13.25−0 cm). Species are 276
grouped into one of three habitat types (planktonic, benthic, or facultatively planktonic) based 277
on classifications by Spaulding et al. [2017] (Fig. 3).
278
Diatom Zone 1 (587−452 cm; ca. 9.6−8.6 ka) is dominated by Staurosirella pinnata 279
(33 %), Cyclotella ocellata (18 %), and other small fragilarioid taxa (60 %) (Fig. 3). By ca.
280
9.0 ka the abundance of S. pinnata decreases to 10 % and the planktonic species Cyclotella 281
rossii (10−30 %), Aulacoseira subarctica (4−25 %) and Cyclotella ocellata (5−14 %) are 282
more dominant. Some of the small benthic species all show slight increases in abundance at 283
this time, including Psammothidium levanderi (9 %) and Achnanthidium minutissimum (6%), 284
albeit at a low relative abundance.
285
In Zone 2 (452−352 cm; ca. 8.6−4.4 ka) the planktonic species C. ocellata, A.
286
subarctica, and C. rossii begin to dominate the assemblage (Fig. 3). Collectively these 287
species reach a maximum abundance of 75 % between 8.5−7.6 ka; a time when small benthic 288
and facultatively planktonic taxa are at their overall lowest Holocene abundances (0−5 %).
289
Increases in abundances of Rossithidium pussilum and other small fragilarioid taxa occur ca.
290
7.6 and 6.8 ka, concurrent with a decrease in planktonic taxa (Fig. 3). After ca. 5.0 ka, the 291
abundance of planktonic species gradually decrease, paralleled by increasing abundance of 292
facultatively planktonic taxa.
293 294
295
Figure 3. Heart Lake diatom stratigraphy and Principal Components Analysis (PCA) 296
scores of the 11 dominant diatom species (>5 % abundance), grouped by habitat 297
preference. Diatom zone demarcation (dashed lines 1−4) is guided by the CONISS 298
cluster analysis. Variables are plotted on a linear timescale (ka BP) and the depth scale 299
refers to depth below lake floor.
300 301
At the onset of Zone 3 (352−13.25 cm; 4.4 ka−1860 AD) a large increase in the 302
facultatively planktonic taxa is paralleled by declines in planktonic taxa (Fig. 3). Collectively, 303
the small fragilarioid taxa make up ~80 % of the assemblages in this zone and several species 304
attain their maximum Holocene abundance, including S. pinnata at 4.2 ka (39 %) and 305
Staurosira construens at 3.8 ka (28 %). In contrast, planktonic species decline from a mean 306
abundance of 55 % in Zone 2, to 5 % in Zone 3. Only Tabellaria flocculosa shows relatively 307
little change in abundance from Zone 2, remaining at ~4%. Of the benthic taxa, Stauroforma 308
exiguiformis and R. pusillum are also present in high abundances throughout Zone 3, with the 309
former attaining a maximum Holocene abundance of 26 % at ca. 2.2 ka.
310
In Zone 4 (13.25−0 cm; ca. 1860−2009 AD) the small fragilarioid taxa continue to 311
dominate the assemblage, comprising ~75 % of the total assemblage ca. 1910 AD (Fig. 3).
312
After this time, the abundance of facultatively planktonic taxa steadily decreases as the 313
benthic and planktonic species increase. After ca. 1970 AD, the numbers of A. subarctica 314
decreases substantially, such that only a few individual frustules were counted per sample.
315
Stratigraphic changes in diatom flora are captured in the first two PCA components, 316
which collectively account for 71 % of the total assemblage variance (Fig. 4). Additional 317
eigenvectors defined by the PCA (3−5) were not considered given they explain progressively 318
lower proportions of the total variance (λ3= 0.108, λ4=0.059, λ5=0.038). PCA 1 represents 57 319
% of total variance and correlates to the planktonic species at the positive extreme, and the 320
facultatively planktonic species at the negative extreme. PCA 2 accounts for 14 % of total 321
variance, and correlates to the small fragilarioid taxa (Fig. 4). The Holocene succession of 322
diatom communities in Heart Lake is further illustrated by the time-series of the 54 sample 323
scores on PCA axis 1 (Fig. 3).
324
325
Figure 4. Loadings of the 11 dominant diatom taxa from Heart Lake and their 326
corresponding PCA scores. Sample scores (circles) are coloured according to their down 327
core diatom assemblage Zone (1−4). Dashed coloured ellipses group diatom species by 328
their habitat preference.
329 330
4.2. Oxygen isotopes 331
Holocene δ18Odiatom values vary between 24.6 ‰ (1805 AD) and 33.3 ‰ (7.6 ka) (𝑥̅ = 29.7 332
‰, n = 137) (Fig. 5) with a range of ±8.7 ‰ that is appreciably greater than the standard 333
deviation of all samples (±0.19 ‰) and diatom standards (±0.30 ‰) measured. The base of 334
the Heart Lake sediment core has a δ18Odiatom value of 29.7 ‰ at 9.6 ka, and values steadily 335
increase to the maximum Holocene value of 33.3 ‰ at ca. 7.6 ka (Fig. 5). After 4.9 ka 336
δ18Odiatom becomes progressively lower until ca. 3.5 ka (27.8 ‰) where values remain stable 337
at ~29−30 ‰ until ca. 1.0 ka. After ca. 1.0 ka, δ18Odiatom exhibits high variability to lower 338
values ca. 1250−1340 AD and 1430−1525 AD, and after 1640 AD there is a shift to overall 339
lower δ18Odiatom values, including the Holocene minimum δ18Odiatom value of 24.6 ‰ at 1805 340
AD. The δ18Odiatom values then slightly increase between 1805−1903 AD, before decreasing 341
to the present day (29.8 ‰) (Fig. 5). Using the sub-division age of 4.2 ka for the mid-late 342
Holocene boundary [Walker et al. 2012], late Holocene δ18Odiatom is significantly (p < 0.001) 343
lower than in the early−mid Holocene.
344 345
346
Figure 5. Time series of Heart Lake δ18Odiatom during (a) the past millennium and (b) the 347
Holocene. Horizontal dashed grey lines indicate the Holocene and the 21st century mean 348
δ18Odiatom value. Orange diamonds and white triangles indicate previously published 349
radiocarbon ages and tephra beds, respectively [Krawiec et al. 2013]. Vertical blue bars 350
correspond to three intervals of Little Ice Age glacier advance in mainland Alaska 351
[Solomina et al. 2015].
352
5. Discussion 353
5.1. Oxygen isotope paleohydrology and paleoclimatology 354
Oxygen isotope ratios measured in precipitation (δ18OP) at Adak airport (1962−67, 1972−73;
355
n = 60) indicate mean annual precipitation-weighted δ18OP is −8.8 ‰, with small seasonal 356
differences between January (–9.4 ‰) and July (–8.9 ‰) [IAEA/WMO, 2017]. The 357
correspondence between Heart Lake water δ18O and the local and global meteoric water lines 358
confirms that (1) Heart Lake water δ18O reflects local precipitation, and (2) evaporative 359
effects influencing precipitation and lake water δ18O are minimal with no isotopic enrichment 360
(Fig. 6). Specifically, the two Heart Lake bottom water (δ18Owater) samples collected in 361
summer 2009 and 2010 (𝑥̅ = –9.5 ‰) are directly comparable, within error, to the long term 362
winter and spring δ18OP values from Adak airport. These data indicate the lake water budget 363
is dominated by winter and spring precipitation (i.e. snowfall and melt) similar to many lakes 364
and streams across Alaska [Clegg and Hu, 2010; Lachniet et al. 2016; Vachula et al. 2017].
365
There is no correlation between mean monthly δ18OP and SAT (r = 0.15, n = 72) or 366
precipitation amount (r = 0.03, n = 72) at Adak airport. Instead, Bailey et al. [2015] found 367
that Adak Island δ18OP values are primarily controlled by the moisture source and trajectory 368
of local precipitating storm systems. Specifically, winters with intensified Aleutian Low 369
circulation are characterized by precipitation with significantly (p < 0.05) lower than mean 370
δ18OP values. These variations are explained by systematic shifts in the central foci of the 371
Aleutian Low; when the SLP minimum is near Adak (strong Aleutian Low), polar air masses 372
are drawn south and advect water vapor and precipitation that is relatively depleted in 18O, 373
along with lower-than-average winter temperatures and increased snowfall (Fig. 2b) 374
[Rodionov et al. 2007; Bailey et al. 2015]. In contrast, a weakened and westerly displaced 375
Aleutian Low increases the southerly Pacific moisture flux to Adak via an enhanced south- 376
westerly storm track (Fig. 2a) [Rodionov et al. 2007]. These systems carry warm 18O- 377
enriched moisture, and bring higher-than-average temperatures and increased precipitation to 378
Adak Island [Bailey et al. 2015].
379 380
381
Figure 6. Heart Lake surface water δ18O (2009 and 2010) on the local meteoric 382
water line (LMWL) and the global meteoric water line (GMWL). LMWL data are 383
derived from Adak monthly composite precipitation samples collected by the Global 384
Network of Isotopes in Precipitation (GNIP) [IAEA/WMO, 2017]
385 386
δ18Odiatom is controlled by several environmental parameters which depend on local 387
hydrology, climate, and the seasonality of diatom growth [Barker et al. 2001; Rioual et al.
388
2001; Jones et al. 2004; Rosqvist et al. 2004; Leng and Barker, 2006; Schiff et al. 2009;
389
Mackay et al. 2011; Meyer et al. 2014; Chapligin et al. 2016]. Previous work by Bailey et al.
390
[2015] showed that the surface core δ18Odiatom record from Heart Lake correlates significantly 391
with the winter NPI during the instrumental period (1900−2009 AD) (r = 0.43, p < 0.02, n = 392
28). This positive relationship confirms that Heart Lake diatoms precipitate their silica 393
frustule in isotopic equilibrium with the lake water in which they grow [Labeyrie, 1974;
394
Leclerc and Labeyrie, 1987], independent of size or species-related vital effects [Bailey et al.
395
2014]. During the spring thaw, it is evident that winter season precipitation (δ18OP) enters 396
Heart Lake coincident with onset of the spring diatom bloom. A limited component of 397
residual summer growth might be expected, but bulk δ18Odiatom analysis is weighted toward 398
the main period of diatom growth in spring [Leng et al. 2001; Bailey et al. 2014]. Under the 399
assumption that similar climatic controls on δ18OP prevailed before 1900 AD, we use this 400
extended δ18Odiatom record as a proxy for atmospheric circulation throughout the Holocene.
401 402
5.2. Holocene environmental history of Adak Island 403
5.2.1. Early-mid Holocene, 9.6 – 4.4 ka 404
Adak Island, along with the Aleutian chain, was glaciated during the last glacial maximum, 405
though there are few chronological constraints on the onset and pattern of ice retreat [Coats, 406
1956; Bradley, 1948; Fraser and Snyder, 1959; Black, 1976]. At Heart Lake, percussion 407
coring ceased at a depth of 587 cm without penetrating bedrock or till, indicating the 408
catchment deglaciated prior to 9.6 ka.
409
From 9.6−9.0 ka, the dominance of fragilarioid and other small benthic taxa reflect a 410
temperate oligotrophic shallow lake with an extensive littoral zone. These pioneering taxa 411
dominate polar to subpolar and mountainous tundra lakes [Lotter and Bigler, 2000; Rühland 412
et al. 2003; Hausmann and Pienitz, 2009; Devlin and Finkelstein, 2011] and their presence 413
suggests a relatively short growth season with cool air temperatures [Smol et al. 2005;
414
Rühland et al. 2008; Hausmann and Pienitz, 2009]. Cool/dry conditions at this time are 415
further supported by low concentrations of biogenic silica (BSi) and organic matter (OM) in 416
nearby Andrew Lake [Krawiec and Kaufman, 2014] and the dominance of Salix and 417
Empetrum in northern Adak [Heusser, 1978].
418
Heart Lake was increasingly colonized by planktonic diatoms between 9.3−4.4 ka 419
(Fig. 3). Of these, A. subarctica is common across Arctic and subarctic zones, and typically 420
shows pronounced periodicity with the spring maximum in non-stratified lakes [Bradbury et 421
al. 2002; Baier et al. 2004; Rioual et al. 2007; Gibson et al. 2003; Solovieva et al. 2015]. It is 422
a heavily silicified species, forming colonies that require turbulence-induced suspension to 423
remain within the photic zone [Rühland et al. 2008; Lotter et al. 2010], and indicates 424
persistent strong seasonal winds, together with associated turbulent water mixing and nutrient 425
upwelling [Wang et al. 2008; Andrén et al. 2015; Solovieva et al. 2015]. In contrast, 426
Cyclotella species have a competitive advantage over the heavily silicified A. subarctica 427
during strong stratification [Andrén et al. 2015] and typically bloom after ice-out in subarctic 428
regions [Rühland et al. 2008; Hoff et al. 2015]. In Kamchatka, Cyclotella spp. prosper during 429
warmer years [Lepskaya et al. 2010], and are broadly considered warm water indicators due 430
to their recent expansion across Arctic lakes [Smol et al. 2005; Rühland et al. 2008].
431
Collectively, these early-mid Holocene diatom assemblages reveal a phase of overall high 432
lake mixing and turbidity, reduced lake ice cover, and relatively high Si/P ratios [Interlandi et 433
al. 1999; Rühland et al. 2003; Rioual et al. 2007]. These changes are further summarized by 434
the Holocene time series of PCA 1 sample scores (Fig. 3).
435
The isotope composition of Heart Lake water was significantly (p < 0.001) higher 436
during the early-mid Holocene compared to the late Holocene (Fig. 5), reflecting the 437
prevalence of southerly storms delivering abundant precipitation with higher δ18O values 438
[Bailey et al. 2015]. Such warm, southerly winter storms would promote turbulent mixing 439
and limit the development of winter lake ice, thereby extending the open-water growing 440
season and allowing for a spring diatom assemblage dominated by planktonic species (Fig.
441
3). Aulacoseira subarctica, in particular, is abundant in modern lake systems during years 442
with short, warm winters [Gibson et al. 2003; Horn et al. 2011]. Elevated pollen percentages 443
of Cyperaceae and other wetland species in northern Adak also imply warm/wet conditions at 444
this time [Heusser, 1978] and correspond to higher local lake levels prior to 4.0 ka [Krawiec 445
and Kaufman, 2014]. Peak δ18Odiatom (33.3 ‰) suggests maximum Holocene warmth at 7.6 446
ka, an inference supported by the simultaneous maximum Holocene abundance of the warm 447
water indicator C. occellata [Rühland et al. 2008] (Fig. 3).
448
The δ18Odiatom record correlates positively with the time series of PCA 1 scores (r = 449
0.48, p < 0.001) and demonstrates that diatom community structure is indirectly connected to 450
climate over millennial timescales. It also indicates that diatom species changes are a natural 451
ecological response to climatically-driven shifts in lake water δ18O, as reflected in the 452
δ18Odiatom record, rather than the converse (i.e. changes in diatom species drive δ18Odiatom
453
variation).
454 455
5.2.2. Mid-late Holocene, 4.4 ka – present 456
At around 4.4 ka, a major shift in diatom composition occurred with marked changes from a 457
predominantly planktonic assemblage to the dominance of small fragilarioid and benthic taxa 458
(Fig. 3). During this transition the relatively warm, deep, and well-mixed open-water 459
conditions of the early-mid Holocene (9.6−4.4 ka) gave way to a less turbulent, potentially 460
shallower lake. This transition coincides with a shift to lower δ18Odiatom values in the late 461
Holocene, reflecting an increase of isotopically depleted water (i.e. snow and/or ice melt) 462
during the spring thaw [Bailey et al., 2015; Streletskiy et al. 2015], and reduced warm, 18O- 463
enriched southerly storms that characterized the early-mid Holocene.
464
An increase in northerly winds and lower temperatures during the late Holocene 465
would have enhanced formation of winter lake ice, which in turn was insulated and prolonged 466
by increased winter snowfall [Mock et al. 1998]. Persistence of lake ice into the spring 467
shortens the aquatic growth season and restricts light penetration into the water column 468
during the spring bloom, thereby precluding the growth and development of planktonic 469
communities requiring an ice-free lake for photosynthesis and a turbulent, well-mixed water 470
column. Instead, the mid-late Holocene flora at Heart Lake is dominated by fragilarioid 471
species known to colonise benthic and periphytic habitats under lake ice cover [Lotter and 472
Bigler, 2000; Douglas and Smol, 2010; Biskaborn et al. 2016]. These benthic communities 473
would have further benefitted from the absence of competition for nutrients from planktonic 474
diatoms, which do not thrive under ice [Lepskaya et al. 2010; Roberts et al. 2015]. A 475
reduction in turbulent wind-driven lake mixing at this time may have also been responsible 476
for increased benthic production and a simultaneous expansion of the littoral zone and 477
benthic habitat [Bradbury, 1988]. Increased winter precipitation and subsequent spring snow 478
melt would account for the sedimentation increase at 3.8 ka from 0.2 to 0.8 mm/yr [Krawiec 479
and Kaufman, 2014]. This turbidity would have further reduced light penetration into the 480
benthic zone, thereby promoting fragilarioid taxa which thrive under limited light and 481
generally turbid conditions [Lotter and Bigler, 2000; Douglas and Smol, 2010].
482
The simultaneous changes in diatom species assemblages and δ18Odiatom values ca. 4.4 483
ka reflect numerous factors affecting vertical mixing patterns, availability of resources (e.g.
484
light, nutrients), and thereby the algal production and composition of Heart Lake. These 485
pronounced changes broadly coincided with other paleoenvironmental changes on Adak 486
Island centred ca. 4.4 ka. For example the BSi and inferred chlorophyll-a record from nearby 487
Andrew Lake also indicates increased aridity after 4.0 ka [Krawiec and Kaufman, 2014], 488
while reconstructed plant assemblages show a reduction in Cyperaceae after ca. 4.5 ka as 489
cooler/drier conditions prevailed over Adak Island [Heusser, 1978].
490
Between 950 AD and 1200 AD, higher δ18Odiatom indicates a transition to overall 491
warmer and wetter conditions on Adak (Fig. 5). A decrease in Empetrum vegetation across 492
northern Adak also indicates increased moisture [Heusser, 1978], while Krawiec and 493
Kaufman [2014] interpret sustained low BSi and chlorophyll-a content from Andrew Lake as 494
the stormiest interval on record. Our δ18Odiatom values exhibit high variability between 950 495
and 1900 AD, implying the local climate was also wetter and more variable since 950 AD.
496
These conditions would account for the continued dominance of fragilarioid taxa over this 497
period with unstable lake conditions [Smol et al. 2005; Rühland et al. 2008; Hausmann and 498
Pienitz, 2009]. Additionally, a peak in sedimentation ca. 1.0 ka, attributed to increased 499
storminess [Krawiec and Kaufman, 2014], rendered conditions unfavourable for planktonic 500
diatom species due to increased sediment suspension and reduced light penetration. Unlike 501
numerous diatom assemblage records across the subarctic and Arctic, in Heart Lake there is 502
no major shift toward those taxa favouring longer growing seasons under warming climatic 503
conditions (e.g. Cyclotella) [Smol et al. 2005]. Conversely, benthic assemblages show an 504
increase after ca. 1860 AD (Fig. 3), reflecting an overall strengthening of Aleutian Low 505
circulation since 1900 AD [Trenberth and Hurrell, 1994] and increasingly unstable 506
environmental conditions on Adak Island over the past century. These findings are consistent 507
with observations in North America and Greenland that suggest shifts in Cyclotella 508
abundances are more closely related to lake mixing, water clarity and resource availability, 509
rather than direct temperature effects [Saros and Anderson, 2015].
510 511
5.3. Regional paleoenvironmental context 512
Our δ18Odiatom reconstruction reveals distinct shifts in the prevailing trajectory of storm 513
systems delivering moisture to Adak Island. The primary trends suggest a relatively weak and 514
westerly positioned Aleutian Low during the early-mid Holocene (9.7−4.5 ka), with a 515
strengthening eastward shift after ca. 4.5 ka (Fig. 5). Based on 21st century observations, 516
typical climatic responses to a weakened Aleutian Low are: (1) a weakening of Pacific mid- 517
latitude storm tracks; (2) increased meridional flow to the central-western Bering Sea; and (3) 518
reduced winter sea surface heat loss in the central-western Bering Sea and enhanced heat loss 519
from the Okhotsk Sea [Mock et al. 1998; Rodionov et al. 2007]. Under this synoptic regime 520
the following conditions would be anticipated in regional paleoclimate records: (1) a 521
reduction in winter storms and precipitation in the Gulf of Alaska region; (2) positive 522
precipitation and temperature anomalies in the central-western Aleutian Islands; and (3) SST 523
warming and reduced winter sea ice extent in the central-western Bering Sea and contrary 524
conditions in the Okhotsk Sea.
525
In support of this synoptic-scale picture, vegetation and lake-level reconstructions 526
provide independent evidence for considerably drier winter conditions in eastern Beringia 527
during the early-mid Holocene [Anderson et al. 2005; RS Anderson et al., 2006; Zander et al.
528
2013]. For example, numerous lakes in southern Alaska and the Yukon record lower-than- 529
present water levels during the early Holocene until ca. 8 ka[Kaufman et al. 2016], reflecting 530
a combination of higher summer temperatures and lower winter precipitation. Furthermore, 531
an inferred decrease in frequency and intensity of winter storms steered into the Gulf of 532
Alaska accounts for marked episodes of glacial retreat at this time[Solomina et al. 2015], 533
driven by reduced winter snowfall/accumulation and negative net mass balance.
534
The SST patterns associated with a weakened wintertime Aleutian Low are also 535
evident during the early-mid Holocene. Relatively warm early Holocene SSTs are 536
documented from the western Bering Sea[Max et al. 2012], reflecting a persistently negative 537
phase of the PDO during the early-mid Holocene and an increase in Pacific storms tracking 538
into the region [Rodionov et al. 2007]. In the Okhotsk Sea, alkenone-derived SST estimates 539
correspond well with Heart Lake δ18Odiatom between ca. 9.6−5.0 ka (Fig. 7), whereby higher 540
δ18Odiatom and an inferred weak Aleutian Low corresponds to lower early-mid Holocene SSTs 541
[Max et al. 2012]. This relationship conforms to modern northerly geostrophic wind 542
anomalies during a weakened and westward displaced Aleutian Low that cool and enhance 543
polynya growth in the Okhotsk Sea [Itaki and Ikahara, 2004; Harada et al. 2014].
544
Specifically, warm (cold) winter SSTs in the Bering Sea (Okhotsk Sea) presently occur when 545
the Aleutian Low is shifted west and the Siberian High dominates over central western 546
Siberia[Rodionov et al. 2007]. These anti-correlated trends also manifest in sea-ice anomalies 547
on weekly to monthly time-scales during the 21st century[Cavalieri and Parkinson, 1987]
548
and are linked to the east–west migration of the Siberian High and Aleutian Low.
549
550
Figure 7. Holocene time series of (a) summer (JJA) insolation at 65°N[Berger and 551
Loutre, 1991], (b) alkenone SSTs from LV29-114-3 in the Okhotsk Sea[Max et al. 2012], 552
(c) Pechora Lake δ18O[Hammarlund et al. 2015], (d) Heart Lake δ18Odiatom (this record), 553
(e) intervals of expanded mountain glaciers in eastern Beringia [Solomina et al. 2015], 554
(f) Mica Lake δ18O[Schiff et al. 2009], (g) Mount Logan ice δ18O[Fisher et al. 2008], (h) 555
Horse Trail Fen δ18O[Jones et al. 2014], and (i) Jellybean Lake δ18O[Anderson et al.
556
2005]. Black lines in (g) and (i) represent 40-yr smoothed intervals. Vertical red shading 557
indicates the eastern Beringia mid-Holocene Thermal Maximum[Kaufman et al. 2016], 558
blue shading indicates the Little Ice Age (LIA) [Solomina et al. 2015].
559
We find independent support for the Holocene migration of the Siberian High from 560
the Pechora Lake δ18Orecord in northern Kamchatka[Hammarlund et al. 2015] (Fig. 7). A 561
north-eastward shift of the Siberian High, concurrent with a strong and eastward shifted 562
Aleutian Low, is linked to periods of increased winter snow contributions to Pechora Lake 563
and overall lower δ18Ovalues[Hammarlund et al. 2015]. The coherency of abrupt and 564
persistent change between the Heart and Pechora Lake δ18O records between 9.6−3.5 ka 565
provides convincing evidence that the Aleutian Low−Siberian High system prevailed 566
throughout the early-mid Holocene (Fig. 7). Moreover, we propose that the synchronous 567
west-east migration of these systems may have been partially responsible for the non-linear 568
and heterogeneous climatic patterns reconstructed across east and west Beringia at this time 569
[Brooks et al. 2015; Kaufman et al. 2016].
570
Maximum values of δ18Odiatom in Heart Lake at 7.6 ka broadly coincide with the 571
northern high-latitude (65 °N) summer insolation maxima ca. 8.0 ka (Fig. 7) [Berger and 572
Loutre 1991]. Significantly (p < 0.001) higher δ18Odiatom in Heart Lake − relative to both the 573
modern (1900 AD−present) and long-term (9.6 ka−present) mean δ18Odiatom – implies a HTM 574
in the central Aleutian Islands at 7.6 ka characterized by persistently weak Aleutian Low 575
circulation, and coincident with maximum abundances of warm water indicator species[Smol 576
et al. 2005] (Fig. 3). Similarly, a Holocene SST maximum is evident ca. 7.5 ka in both the 577
northwest Pacific[Minoshima et al. 2007] and the subarctic North Pacific[Harada et al.
578
2014], and from GCMs which indicate maximum SATs and SSTs in the Bering Sea and 579
Aleutian Islands ca. 7.0−8.0 ka[Renssen et al. 2012]. In southern Kamchatka, the majority of 580
paleoenvironmental records demonstrate a HTM ca. 7.0−5.3 ka[Brooks et al. 2015], 581
consistent with warm temperatures across eastern Beringia[Kaufman et al. 2016]. These 582
results contrast with previous paleoclimate studies from Alaska and the northwest Pacific that 583
identify an earlier HTM at ca. 11.3−9.1 ka [Kaufman et al. 2004; Max et al. 2012]. Such 584
uncertainty in these early Holocene warming patterns is highlighted by Zhang et al. [2017]
585
who found large discrepancies between modelled and reconstructed Holocene temperatures 586
across Alaska. Hence, it is difficult to fully constrain the timing of the HTM in the central 587
Aleutian Islands, particularly given that our record does not extend the full Holocene epoch 588
coupled with a paucity of local alternative studies.
589
Simultaneous shifts in diatom flora and δ18Odiatom after the HTM at ca. 4.5 ka indicate 590
multiple and inter-related environmental changes that impacted Heart Lake. These 591
pronounced changes coincide with local proxy inferences demonstrating increased aridity 592
under a prevailing northerly circulation pattern [Heusser, 1978; Corbett et al. 2010; Krawiec 593
and Kaufman, 2014]. This mid-Holocene perturbation coincides with a return to cooler 594
conditions, increased winter precipitation and extensive glacial advance in Kamchatka 595
[Nazarova et al. 2013; Barr and Solomina, 2014; Meyer et al. 2015]. Widespread cooling is 596
also evident in eastern Beringia during the late Holocene[Kaufman et al. 2016], and 597
mountain glaciers across Alaska advanced between ca. 4.5 and 3.0 ka[Solomina et al. 2015], 598
in phase with those in Kamchatka and demarking onset of the Neoglacial across Beringia 599
[Savoskul, 1999; Barr and Solomina, 2014]. Though temperature is proposed as the principal 600
control on regional glacier mass balance through the Holocene [Solomina et al. 2015], the 601
observed glacial maxima in Alaska are asynchronous with the timing of pronounced cold 602
intervals[Kaufman et al. 2016]. Instead, our data suggest the transition to intensified Aleutian 603
Low circulation after 4.5 ka, coincident with declining summer insolation [Berger and 604
Loutre, 1991], drove widespread Neoglacial advance through the combined effect of 605
increased winter snowfall under a generally cooler regime, yielding a marked regional 606
positive mass balance perturbation. In particular, we note during the past millennium three 607
intervals of lower δ18Odiatom values between 1275−1350 AD, 1400−1550 AD, and 1700−1850 608
AD coincide with three well-documented episodes of Little Ice Age (LIA) glacier advance on 609
mainland Alaska (Fig. 5 and 7) [Calkin et al. 2001; Solomina et al. 2015]. Furthermore, the 610
δ18Odiatom minimum at 1805 AD (+24.6 ‰) marks the culmination of regional LIA glacial 611
advance [Barclay et al. 2009; Calkin et al. 2001; Wiles et al. 2004; Solomina et al. 2015]
612
(Fig. 5 and 7).
613 614
5.4 Paleoisotopic coherence and atmospheric circulation 615
Several paleoisotope records from Alaska have also been interpreted in terms of synoptic- 616
scale changes in atmospheric circulation and inter-comparison with Heart Lake δ18Odiatom
617
yields many commonalities and insights [Anderson et al. 2005; Fisher et al. 2004; 2008;
618
Schiff et al. 2009; Jones et al. 2014; Hammarlund et al. 2015] (Fig. 7). For instance, a strong 619
inverse relationship ca. 9.5−4.0 ka is apparent with millennial scale δ18Odiatom variations at 620
Mica Lake, in Prince William Sound[Schiff et al. 2009] (Fig. 7). Lower Mica Lake δ18Odiatom
621
values indicate precipitation delivered by zonal flow under a weak Aleutian Low, whereby 622
precipitating systems are subject to increased rainout as they pass over the Kenai Peninsula 623
and coastal mountain ranges. Conversely, increased meridional flow during a strong Aleutian 624
Low delivers locally sourced moisture from nearby Gulf of Alaska, thereby reducing 625
distillation and isotope depletion in precipitation, thus yielding higher Mica Lake δ18Odiatom
626
values [Schiff et al. 2009]. The reciprocal relationship between precipitation-inferred δ18O 627
values at Heart and Mica Lakes between ca. 9.5−4.0 ka conforms to modelling and empirical 628
analyses of spatial patterns of δ18OP [Berkelhammer et al. 2012; Bailey et al. 2015]. The 629
Horse Trail Fen record from the Kenai lowlands is also comparable to Heart Lake from ca.
630
8.0 ka and demonstrates overall higher δ18O values during the early Holocene and reflecting 631
generally weak Aleutian Low circulation [Jones et al. 2014]. The only other full Holocene 632
paleoisotope record from eastern Beringia is from the Mount Logan ice core [Fisher et al.
633
2008], which exhibits strong correspondence with the Jellybean [Anderson et al. 2005] and 634
Heart Lake δ18O records during the early-mid Holocene (Fig. 7).
635
Secondary, but notable departures between paleoisotope records are evident during 636
the late Holocene(Fig. 7), some of which can be reconciled by considering the detailed, non- 637
